In my review of the Nikon D800 I have a number of pages on autofocus, including one page that shows inconsistent results with the left/right AF sensors.
Geoff B writes:
I have followed your testing of the D800, and am impressed with your rigour and fearlessness in testing. You have posted some tests on the AF system with various lenses. I am assuming you used the central AF point to do this testing. Have you encountered the left side AF focus inaccuracy in any testing? Here is a thread from Nikonians that attempts to quantify the issue.
This is a separate issue to the general focusing problems that you have identified over the years, most notably with the new G primes. I am starting to despair that Nikon can get this AF system of theirs working reliably.
On the linked thread at the time of my writing there are 24 individuals who have identified this fault in their D800 out of a sample size of 65; ie, 37%. I bought and returned a D800 on 25th May that was badly misfocusing on the left AF point on most of my Nikon lenses (not the 105G VR). What do you make of this. I presume its another example of poor QC, or the demands of 36MP not matching manufacturing tolerances available for the price. What is it about DSLR autofocus that is so difficult?
DIGLLOYD: I have not yet evaluated my D800E for autofocus; it’s a tedious boring job I have been putting off!
In general, off-center focus is more subject to errors because there tend to be a greater level of optical aberrations, which make it increasingly difficult for the AF system to distinguish contrast levels.
Some caution is advised here, since field curvature exists with fast lenses. Sharpness symmetry is an issue when evaluating focus; not having the center and mid-zone and edge sharpness the same is normal (one must understand the field curvature of the lens in use. The issue is whether one gets the same result using off-center AF points, in the sense of left/right/center consistency for left vs right AF sensor.
One must also not rule out lens decentering or other issues, therefore all tests should also be done with manual focus as a baseline so that lens issues are not misinterpreted as autofocus issues.
I have received my D800 over a month ago, and as one would guess, it's murphy's law that this unit showed bad back focus with the left and right AF points (worst on the left side), but perfect at the centre. Nikon then issued me with a replacement, but while I was waiting for it, they collected my original unit and supplied me with a loan unit. Ironically, this unit also had the same AF issue. The new replacement unit I received was personally checked by a technician for this AF issue prior to sending it to me, and indeed it showed even AF performance across the board. However, all with a slight front focus. No big deal. Here comes the but... both the loan unit and the new replacement unit showed a very strange issue in the viewfinder, something I doubt many people would/could see. Right down the centre, a vertical blurry band could be seen. Running down the centre and through the centre AF point, but a little narrower in width than the AF point itself.
This problem is a great source of headache when using manual focus lenses and a DK-17m eye piece attached. This blur band is most noticeable with my Distagon 35/2 and AF-D 85/1.4 lenses (for reasons unknown to me), and the best subject matter would be well lit, very fine detailed subject matter. For example, tree leaves very far away, a field of canola. Very high frequency detail.
With this, Nikon actually issued another replacement for me, as well as send my original unit (AF calibrated) to the local Nikon office, and told me to make a choice.
I picked up my original unit and checked for focus error. The uneven performance was still there, but the blurry smudge was not. I then went to the very new unit they sent down, and low and behold, it's pentaprism smudge was even worse. To my surprise, the Nikon technician saw it as well, and told me the unit was not right. This unit showed no uneven AF performance.
So now, I'm back to square one after 1 month, with the same camera and the same uneven AF issue. I obviously don't mind that I'm back to unit 1, but I have an issue that the focus problems have not been fixed, even though the report card shows that work had been done to calibrate the points.
Anyways. The point is, out of the four D800 cameras I've handled so far, three of them showed the pentaprism blur/smudge down the centre. Out of the four cameras, two showed the uneven AF performance (Left back-focuses, centre is perfect, and right back-focuses again but not as bad as the left side).
DIGLLOYD: Ouch. See comments below for the explanation.
Panikos H writes:
May I start by adding to the compliments on your thorough and unbiased evaluation for everything you test. I do find it so refreshing!
Regarding Martins (D800 autofocus - 2 June 2012) comment on the "vertical blurry band", this is not an unknown issue. It is something I experienced from time to time while I supervised Nikon's UK service department. It is where the pentaprism isn't correctly "aligned" when the cast is made which creates this rather disturbing blur effect. It has no adverse effect on images, just on what is viewed when looking through the pentaprism. The "alignment" in question is the angle at which the two planes of glass meet at the top of the prism, the very top edge in the attached image.
The only remedy is to replace the prism with a new one. On some of the newer and lighter cameras, the pentaprism isn't a piece of solid glass that is then silvered on 5 of its 7 sides, but a number of mirrors that are stuck together to form the same shape & effect as a pentaprism, with a thin piece of glass at the bottom. So what you have is a "hollow" but lightweight pentaprism. It is on these lightweight units that this vertical blurry line is most common, as it is possible for the top (side) planes to be incorrectly stuck together during the manufacturing process.
I hope this helps, and thank you again for your excellent articles.
DIGLLOYD: Very interesting indeed!
Thorseten K writes:
I feel for for your readers who reported the D800 AF issues. I went thru 4 (!) D800 bodies before I found one with no AF problem.
The first 3 units had the widely reported left AF issue. It was very obviously back-focusing with the left-most AF focus point, when using f/1.4 glass or wide angle zooms and shooting a subject 10ft or more away. It also back-focused under other conditions (with other points left of center and closer focus distances and other lenses) but to a lesser degree.
I sent the first body to Nikon service. It came back "AF repaired and adjusted" but still had the same issue, except it now also had dust on the sensor and viewfinder focus screen. I returned it to the dealer. The 2nd body had the same issue. I didn't bother with Nikon and brought it back to the dealer right away. The dealer said they were expecting a visit from Nikon (one technician and a rep) the next day, and he would suggest letting them test my camera to make sure it really has an issue. They were probably getting suspicious.
Well guess what, the tech came, took a look and said, yes, it was wrong and needs to go back to Nikon. So I didn't imagine things.
Body number 3 - same issue. Then I waited a month, and yesterday finally I got a body which looks to be focusing properly with all AF points. It does have back focus with some of my lenses (especially wide angle), but it's even across the points so it can be adjusted with AF fine tune. So it looks they may have sorted out the issue now, but Nikon's quality control certainly didn't impress me with the D800 release.
DIGLLOYD: I suppose once things settle down I’ll have to send my D800 and D800E in to be checked. The D800 is off, but I have not yet tested the D800E autofocus.
Scott M writes:
Thank you (and Panikos H!) for the brief rundown on the vertical blurry line running through the VF. I have this irritating problem on my D700 and if it were not offset slightly to the left (as it is it lies slightly outside the center AF rectangle), MF would be virtually impossible without resorting to the clunky implementation of LV on this camera. It's worse on some subjects than others, being especially visible on distant foliage.
I posted about this on the DPR forums and while numerous folks experienced it to one degree of another I never got to the bottom of it (I assumed it was due to some defect in the roof of the prism - I just never knew they were moulded, I always thought they were ground and polished from a block).
Weirdly I have never seen this effect even on inexpensive film SLRs from way back. Maybe glass moulding was not up to the task back then and we had to settle for ground and polished prisms?
DIGLLOYD: Seems to me that a solid piece of glass is always better, and I suspect that this was the case with older cameras and perhaps with the Nikon/Canon pro bodies (but I don’t know).